On Thu, May 16, 2002 at 10:26:46AM -0500, Stephen Lord wrote:
> On Thu, 2002-05-16 at 10:17, Andrea Arcangeli wrote:
> > On Thu, May 16, 2002 at 09:50:31AM -0400, Jim Eshleman wrote:
> > > After two weeks of uneventful uptime with -pre6aa1 the system hung
> > > with no symptoms, no messages, no way in. XFS filesystems recovered
> > > themselves on reboot with no apparent problems. About twelve hours
> > > later got the appended oops, in the early AM. I suspect it may have
> > > been triggered by the system backup, which is now stuck in D state.
> > > Otherwise the system is running normally except loadavg is high (6-7)
> > > even though there is little CPU and IO usage.
> >
> > I would suggest a full forced fsck and an upgrade to 2.4.19pre8aa3 that
> > includes the xfs 1.1 release (it fixes various xfs issues compared to
> > 2.4.19pre6aa1). Personally I mostly care about the glue between xfs and
> > mainline kernel, so for specific xfs bugs you were right to CC the xfs
> > mailing list.
> >
> > Andrea
>
> Well with xfs there is no fsck, but there is xfs_repair which is
> the equivalent. You can also run xfs_check which will not fix a
> filesystem but runs an extensive set of checks on its consistency.
Yep.
>
> Andrea the interfaces between xfs and the kernel are about to change,
> we will no longer require code in vmscan.c and the code in buffer.c
> gets simpler.
>
> I was not aware that the xfs code in the aa kernels still
> predated 1.1, in that case upgrading is definitely a good idea.
the reason is that at the time 2.4.19pre6aa1 was released xfs 1.1 wasn't
yet released.
Andrea
|